On 22/03/2008, Kirk <[EMAIL PROTECTED]> wrote: > rob hinds wrote: > > Hi All, > > > > I have been running some performance tests using jmeter (testing > > webservices) and the results i am receiving do not seem consistent. > > > > I am running http tests and writing the results to a csv file, storing the > > timestamp and latency of each test. > > > > I am calculating the overall time of the test run(the timestamp of the last > > test minus the timestamp of the first test) and then calculate the thruput > > by dividing the number of tests run (I do a run of 2500 requests) by the > > calculated time taken. > > > > however, the thruput does not seem to be consistent with the average > > latency. > > > > For example, I run a test of 2500 and the average latency is 500ms and the > > thruput is 15.1 messages/second, however, I also run another test and I get > > an average latency of 250ms but the thruput is only 17.4 messages/second. > > Intuitively, the latency should directly proportional to the thruput, and > > roughly speaking (untill you max out your services) if you double the > > average latency I would expect the thruput to halve? > > > > Any ideas as to what might be happening? > > > > what hardware are you using? How is the test setup? It is possible that > you've saturated the system in such a way that the only thing you've > done by reducing the load is you've reduced the latency. IOWs, you're > beyond the knee >
The latency is time to first response. If there are a lot of large pages, that may not scale. > Kind regards, > Kirk Pepperdine > > Thanks in advance > > Rob > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
